{"id":55027,"date":"2026-01-14T14:12:40","date_gmt":"2026-01-14T06:12:40","guid":{"rendered":"https:\/\/ljdevice.com.tw\/?p=55027"},"modified":"2026-01-14T14:12:40","modified_gmt":"2026-01-14T06:12:40","slug":"exclusive-nvidia-turning-flash-ai-memory","status":"publish","type":"post","link":"https:\/\/ljdevice.com.tw\/en\/exclusive-nvidia-turning-flash-ai-memory\/","title":{"rendered":"Exclusive: Why Nvidia Is Turning to Flash for AI Memory"},"content":{"rendered":"

2026-01-12
\n<\/p>\n

\n
\n
\n
\n
\"Exclusive:<\/p>\n

Source\uff1aTechTaiwan<\/a><\/p>\n

At CES 2026, Nvidia revealed a major shift in AI storage\u2014offloading long-term \u201ccontext memory\u201d from HBM to flash-based SSDs. A surprise cameo by Phison\u2019s chip in Nvidia\u2019s new server sparked market buzz, but CEO K.S. Pua remains cautious about its real impact.<\/p>\n<\/section>\n

\n
\n

Early Jan. 7 morning, a photograph began circulating among stock-market traders,\u00a0showing a NAND flash controller IC from Phison inside Nvidia\u2019s newly unveiled Vera Rubin server at CES.<\/strong><\/p>\n

This widely shared \u201cbullish signal\u201d was one of the factors that helped drive Phison shares to their daily limit that day, before a sharp correction on Friday.<\/p>\n

When I first saw the image, I immediately thought of something Phison CEO K.S. Pua told me more than two months ago:\u00a0\u201cIn the future, flash memory will become part of AI computing systems\u2014and when that happens, Phison will be part of it.\u201d<\/strong><\/p>\n

Had that prediction really come true so quickly?<\/p>\n

In Jensen Huang\u2019s one-hour-and-40-minute keynote at CES, the final ten minutes were devoted to the launch of Nvidia\u2019s fourth-generation BlueField data management system.<\/p>\n

One of its key innovations aligns precisely with a direction Phison has invested heavily in over the past several years:\u00a0offloading part of the storage burden from high-bandwidth memory (HBM) to flash-based solid-state drives (SSDs).<\/strong><\/p>\n

What gets offloaded, Huang explained, is so-called \u201ccontext memory.\u201d<\/p>\n

\u201cThat\u2019s the accumulated history of your conversations with AI,\u201d he said. \u201cOver time, you\u2019ll want to preserve those memories forever, right? Initially, that context memory sits in HBM\u2014but eventually, it simply won\u2019t fit anymore.\u201d<\/p>\n

The solution, he said, is to move those hard-to-discard memories to a separate storage rack and retrieve them via high-speed networking.<\/p>\n

Pointing to the rack on stage, Huang said this approach effectively provides an additional 16TB of storage per GPU.<\/p>\n

In a post-keynote earnings call, he elaborated further, describing this new storage architecture\u2014purpose-built for AI key-value (KV) caches and context memory\u2014as something that \u201cwill create a market that has never existed before, and could become the largest storage market in the world.\u201d<\/p>\n

Several U.S. media outlets noted that it was this \u201clargest in the world\u201d remark that helped ignite this week\u2019s sharp rally in memory-related stocks.<\/p>\n

If even part of this new storage demand were to use Phison\u2019s controller ICs, wouldn\u2019t that be a massive windfall for the company?<\/p>\n

But when Phison CEO K.S. Pua\u2014who was attending CES\u2014returned my call on 8th Jan, he answered in his usual blunt fashion: no.<\/p>\n

Source:Exclusive: Why Nvidia Is Turning to Flash for AI Memory\uff5cIndustry\uff5c2026-01-12\uff5cweb only<\/a><\/p>\n

<\/div>\n<\/div>\n<\/section>\n<\/div>\n<\/header>\n<\/div>\n<\/div>\n
\n
\n
\n
<\/div>\n<\/section>\n<\/div>\n<\/div>\n","protected":false},"excerpt":{"rendered":"

2026-01-12<\/p>\n","protected":false},"author":7,"featured_media":54422,"comment_status":"closed","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[609],"tags":[],"class_list":["post-55027","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-industrial-news"],"_links":{"self":[{"href":"https:\/\/ljdevice.com.tw\/en\/wp-json\/wp\/v2\/posts\/55027","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/ljdevice.com.tw\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/ljdevice.com.tw\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/ljdevice.com.tw\/en\/wp-json\/wp\/v2\/users\/7"}],"replies":[{"embeddable":true,"href":"https:\/\/ljdevice.com.tw\/en\/wp-json\/wp\/v2\/comments?post=55027"}],"version-history":[{"count":1,"href":"https:\/\/ljdevice.com.tw\/en\/wp-json\/wp\/v2\/posts\/55027\/revisions"}],"predecessor-version":[{"id":55028,"href":"https:\/\/ljdevice.com.tw\/en\/wp-json\/wp\/v2\/posts\/55027\/revisions\/55028"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/ljdevice.com.tw\/en\/wp-json\/wp\/v2\/media\/54422"}],"wp:attachment":[{"href":"https:\/\/ljdevice.com.tw\/en\/wp-json\/wp\/v2\/media?parent=55027"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/ljdevice.com.tw\/en\/wp-json\/wp\/v2\/categories?post=55027"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/ljdevice.com.tw\/en\/wp-json\/wp\/v2\/tags?post=55027"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}